A laser beam starts at one edge of the grid, and you have targets it needs to hit. Your job is to place the pieces — mirrors to bend the beam, splitters to divide it, portals to teleport it, blockers to stop it — so the light routes through and lands on every target. Get the path right and the whole board lights up. It's part optics, part puzzle, and entirely about thinking in lines and angles.
A fresh daily puzzle lands each midnight. Send a friend the exact layout with the challenge link.
Placement order matters less than commitment: the beam re-fires instantly with every piece you set, so the game quietly teaches you to read reflections two mirrors ahead. Solutions save nothing to any server — progress lives in your browser. The daily maze is one fixed layout for every player. Path-logic fans should also try Weave ZAP, where you are the beam.
Mirror Maze runs on a decision I'd happily defend against any performance review: the beam re-fires instantly on every single placement. Put a mirror down — the light recalculates and redraws its full path, that very moment. Rotate a piece — same thing. There's no "run" button, no test phase, no commit step. The board is always live.
That choice has a real cost. Recomputing the entire beam path on every move means the engine traces every reflection, every split, every dead-end on each interaction, rather than once per attempt. But I watched what it does for players learning the game, and it isn't close. When the consequence of a placement appears the instant your finger lifts, the reflection rules teach themselves. Nobody reads an optics tutorial; you put a mirror at 45 degrees, watch the beam snap sideways, and your brain files the rule permanently. A run button turns that same lesson into a slow loop of guess, test, forget. Instant feedback is the tutorial. That's why the expensive version won.
Now, the mistake that stalls more Mirror Maze players than any other, and it's a subtle one: placing splitters in the wrong order. A splitter is the most powerful piece in the game — one beam in, multiple beams out — and players understandably reach for it early, planting splitters at every junction where they'll eventually want the light divided. Here's the problem. A splitter only acts where the beam already reaches. It's not a promise about future light; it's an operation on present light. A downstream splitter sitting in darkness does exactly nothing, and because it does nothing visibly, players assume the piece is broken rather than unpowered. I've read that exact support message more than once.
The discipline is: route first, then split. Trace your main beam to the first junction using mirrors alone, confirm the light actually arrives, and only then place the splitter and watch the beam divide. Now walk each branch outward the same way — route it, then split it — building the light-tree from the trunk toward the leaves. Working leaf-first feels efficient and produces a board full of dormant glass.
The live beam makes this discipline pleasant rather than tedious, because every routing step gives you an immediate, visible win — the light creeps further across the board with each correct mirror, and a wrong angle announces itself instantly instead of hiding until some final test run. Follow the light, not your plan for the light.
The daily maze resets at midnight, the same layout for every player. Watching a finished board light every target at once is, I maintain, one of the best single frames anywhere on this site.
Don't start from the laser — start from the targets. I work backwards: a target needs the beam arriving from some direction, which means a mirror has to sit at a particular spot turned a particular way, which means the beam has to reach that spot somehow. Tracing from the destination back toward the source narrows your options far faster than pushing the beam forward and hoping it lands.
When the puzzle hands you a splitter, treat it as a gift, not a complication. One beam becoming two means you can hit two targets from a single source — so the elegant solutions usually run the light through the splitter early and fan it out, rather than trying to snake one thin beam past everything. Think about where the light needs to end, then build the path to deliver it there.
Mirror Maze trains you to think backwards from a goal and to reason about cause and effect in a chain — place this, and three steps later the light arrives there. It's spatial logic with a satisfying click when it resolves. Mirror Maze is best solved backwards — start from where you want the light to land and trace the path back to the source. And honestly, that's how life makes sense too: looking back, the bends and detours and the tricky bits all turn out to have led somewhere, even the ones that felt like setbacks at the time. Working backwards from the outcome, you see that nothing was wasted. Every mirror was placed for a reason, even the ones you'd have skipped if you'd known.
That reverse-planning instinct is genuinely powerful: a lot of hard problems get easier when you start from the outcome and work back to the first move. The puzzles grow more pieces so the chains get longer and the planning deeper. No download, no sign-up — play in your browser on any device. If you like spatial puzzles, try Weave ZAP next.